Re: [libvirt] PATCH 1/4: More generic MAC address handling
"Daniel P. Berrange" <berrange(a)redhat.com> wrote:
...
> Also, unless there's a guarantee that the random number state
is
> initialized elsewhere, it should be initialized here, like it was in
> the now-removed xenXMAutoAssignMac function.
>
> srand((unsigned)time(NULL));
>
> Or maybe just initialize it once at start-up?
Could just add a call to srand() in virInitialize() i guess, although is
that a reasonable thing for a shared library to be doing, rather than
leaving it upto the application ?
You're right that it's not reasonable to call srand from library code.
One alternative is to use something like coreutils' gnulib-style randint
module. I'm looking into whether we can relax it's GPL license to LGPLv2+.
Using it or any other library-safe random_r-like interface would mean
maintaining and reusing an internal-to-libvirt random state buffer.